一种改进的基于自动形态学的端元提取算法

发布时间：2018-05-15 01:26

本文选题：遥感 + 高光谱图像　；参考：《激光技术》2017年01期

【摘要】：自动形态学端元提取(AMEE)算法中的形态学算子在纯像元集中分布的区域无法得到正确的结果。现有膨胀操作在每个结构元素内只能提取一个候选端元,会造成重要像元丢失。为了解决这些问题,采用改进的形态学算子和结构元素对AMEE算法进行了改进。首先引入参考光谱向量的概念构建了改进的形态学算子,并给出了形态学离心率指数新的计算方法,然后利用偶数大小、改进的结构元素,从每个结构元素内选出4个候选端元,最后对改进的基于自动形态学的端元提取算法进行了分析和实验验证。结果表明,改进的方法能从纯像元集中分布的区域获得正确的候选端元,并在一定程度上避免膨胀过程中的信息遗失,从而能够有效地提升端元提取的精度和像元解混的效果。
[Abstract]:The morphological operators in the automatic morphological End-element extraction (AMEE) algorithm can not get the correct results in the region of the pure pixel set distribution. The existing expansion operation can extract only one candidate end element within each structure element, resulting in the loss of important pixels. In order to solve these problems, improved morphological operators and structural elements are used to improve the AMEE algorithm. In this paper, the concept of reference spectral vector is introduced to construct the improved morphological operator, and a new method to calculate the morphological centrifuge index is given. Four candidate endpoints are selected from each structural element. Finally, the improved algorithm based on automatic morphology is analyzed and verified by experiments. The results show that the improved method can obtain the correct candidate endelements from the region of the pure pixel set, and to some extent avoid the loss of information in the expansion process, which can effectively improve the accuracy of the end element extraction and the effect of pixel unmixing.
【作者单位】：杭州电子科技大学自动化学院;中国科学院长春光学精密机械与物理研究所航空成像与测量技术研究部;
【基金】：国家自然科学基金资助项目(61375011) 浙江省自然科学基金资助项目(LY13F030015)
【分类号】：TP751

【相似文献】